The Claude Delusion

It argues that humans reflexively impute intention to things - paintings, conversations, even sunsets when framed as deliberate - and that this reflex makes interactions with AI uncanny and misleading. Using the sunset-versus-painting analogy and Mark Fisher’s definition of eeriness, the argument distinguishes artifacts created by minds from outputs “extruded” by statistical, theory-free systems. AI-generated text and images, no matter how striking, lack an authorial mind; when people treat chatbots as interlocutors they are hallucinating a person into behavior that is actually produced by pattern prediction. Those hallucinations, the piece contends, are more consequential than the technical “errors” usually labeled as AI hallucinations.

Experience with AI diminishes the initial wonder: novelty grades the output on a curve, and repeated exposure reveals a characteristic smoothness and banality where texture and idiosyncratic trace elements should be. AI can automate rote feedback and correct common beginner mistakes, but it cannot perceive or cultivate the subtle, nonstandard features that make expert judgment and unique voices valuable. As familiarity grows, many people are becoming “AI atheists,” recognizing the lack of intent, yet two barriers persist: general unfamiliarity with how tasks are produced makes people assume intent, and absence of domain expertise leaves many unable to detect the machine’s lack of understanding.